Just wanted your thoughts on crossing high (over 5k) vs crossing low (1-1.5k).
For simplicity's sake lets say that we don't want to crossover in 2k-5khz due to the ears sensitivity in this region.
I always thought crossing At 1khz or lower was the holy grail, until i read this information.
"The below pics show that at high frequencies the direction of sound is detected simply by amplitude (loudness) difference between our ears.
At low frequencies the wave lengths are so long that both ears will hear sound at similar level. Therefore below 1K Hz the direction of sound is by the phase difference in the wave between our ears. We are able to detect sound direction to approx 100Hz in a free field. But in a reverberant cave or room this is limited to approx 300Hz because the bass energy is reflected from all directions."
Obviously, crossovers create phase shift. So is it actually better to cross over above 5k rather than below 2k?
For simplicity's sake lets say that we don't want to crossover in 2k-5khz due to the ears sensitivity in this region.
I always thought crossing At 1khz or lower was the holy grail, until i read this information.
"The below pics show that at high frequencies the direction of sound is detected simply by amplitude (loudness) difference between our ears.
At low frequencies the wave lengths are so long that both ears will hear sound at similar level. Therefore below 1K Hz the direction of sound is by the phase difference in the wave between our ears. We are able to detect sound direction to approx 100Hz in a free field. But in a reverberant cave or room this is limited to approx 300Hz because the bass energy is reflected from all directions."
Obviously, crossovers create phase shift. So is it actually better to cross over above 5k rather than below 2k?
I was poking around with a vaguely similar question a little while time ago. As ever there were different viewpoints, but crossing high was felt by some to increase beaming from the mid/mid-bass too much (more of an issue for off-axis listening), whilst conversely others felt that crossing low increased the distortion of the tweeter too much for their taste.
I can't remember who, but someone in another thread pointed out that as the excursion (and therefore distortion) of the tweeter increases 'so much' down towards its fs, there can still be increasing excursion even if theres a crossover attenuating the spl. So maybe one should consider steepness of the XO as well as how far above the tweeter's Fs it is.
Other questions like phase shift and sensitivity vs frequency of human hearing are also considerations - its a huge bundle of compromises between all sorts of competing requirements. I'm too much of a newbie to put these into great perspective but my own personal feeling is that phase shift is perhaps less important than distortion, and whilst i'd like to cross over outside the most sensitive hearing range that hasn't been compatible with the FRs of the driver sizes that I wanted to use.
As I say, I'm a relative newbie so i'd be keen to hear where this discussionm goes and how far off the mark i am.
Cheers
kev
I can't remember who, but someone in another thread pointed out that as the excursion (and therefore distortion) of the tweeter increases 'so much' down towards its fs, there can still be increasing excursion even if theres a crossover attenuating the spl. So maybe one should consider steepness of the XO as well as how far above the tweeter's Fs it is.
Other questions like phase shift and sensitivity vs frequency of human hearing are also considerations - its a huge bundle of compromises between all sorts of competing requirements. I'm too much of a newbie to put these into great perspective but my own personal feeling is that phase shift is perhaps less important than distortion, and whilst i'd like to cross over outside the most sensitive hearing range that hasn't been compatible with the FRs of the driver sizes that I wanted to use.
As I say, I'm a relative newbie so i'd be keen to hear where this discussionm goes and how far off the mark i am.
Cheers
kev
Hi,
It depends on a lot of things, there is no simple answer.
In most cases you can't cross higher than 5KHz due to mid
limitations and lower than 2KHz due to tweeter limitations.
The obvious way is to go for mid/treble unit, a FR, with possibly
a helper tweeter at the top end and low x/o point to the bass.
So called "FASTs" can make very cost effective diy speakers.
rgds, sreten.
It depends on a lot of things, there is no simple answer.
In most cases you can't cross higher than 5KHz due to mid
limitations and lower than 2KHz due to tweeter limitations.
The obvious way is to go for mid/treble unit, a FR, with possibly
a helper tweeter at the top end and low x/o point to the bass.
So called "FASTs" can make very cost effective diy speakers.
rgds, sreten.
This is a point that many here don't seem to understand. It will not only affect the on-axis response, but also the in-room power response, causing all kinds of frequecy response issues. But crossing too low can drastically increase distortion, especially for tweeters, it can be one of the more audble and annoying forms of distortion.
Not only do crossovers cause phase shift, but so do the drivers, the key to success is getting them all to play nicely together.
I think this may have been mentioned before; speaker building is as much Art as Science, and a lot of trade offs are inevitable. Gaining experience, learning about all the things that go into this hobby can seem daunting at times, but when you finally get it right, it is an awsome experience.
Mike
I'd say cross where the drivers change radiation pattern such that you can have constant dispersion over the whole range.
If phase shift is such a problem then you can just buy a MiniSHARC and correct the shifts. When you are at it you could fix the low frequency roll off shift too.
Some thoughts on the specific question:
- Well-designed crossovers result in a phase that shifts smoothly as frequency rises, not a sudden error. Also, there are a handful of ways to end up with zero phase shift around the crossover.
- Both of your speakers hopefully have the same phase characteristics, whatever that may be. Seems to me that this eliminates the problem with localization being described.
- Well-designed crossovers result in a phase that shifts smoothly as frequency rises, not a sudden error. Also, there are a handful of ways to end up with zero phase shift around the crossover.
- Both of your speakers hopefully have the same phase characteristics, whatever that may be. Seems to me that this eliminates the problem with localization being described.
I don't understand why after reading the link you have to change your mind.
The above ideas about crossing below 1K is only a guideline. In the real implementation you may find an optimum number.
Of course you can set your objective prior to that. For example, you want to cross below 1K so you then find suitable tweeter for that purpose (this is expensive). And this approach has merit.
But I can't understand about setting a preliminary objective of crossing at 5KHz so then to find a woofer that can go that high. For sure it will be a small driver and for sure it will be a 3-way to get the bass performance. And this has been done many times with 3-way.
Finding the crossover "sweet spot" for a 3-way, that is an interesting research. I haven't found many information on the net.
Depending on the Woofer and Tweeter, the directivity changes as you change the frequency response. Here I have a 12" woofer and 1" compression driver with waveguide. From measurement the polar response alone 1khz is best, but then we tradeoff with SPL. So 1.5khz is chosen for the best compromise.
At 2.5khz:
At 1.5khz:
At 1khz:
At 2.5khz:
An externally hosted image should be here but it was not working when we last tested it.
At 1.5khz:
An externally hosted image should be here but it was not working when we last tested it.
At 1khz:
An externally hosted image should be here but it was not working when we last tested it.
What sreten says is very true. Myself, i like my 2-way XOs in the 150-350ish Hz range. There are a groing number of drivers that can handle the HF job
dave
Seems like everyone is interpreting this question as just another attempt to make a foolish generalization, and explaining all (or some) of the variables involved in the compromise. I don't think that's fair to the OP. Yes, where to cross depends on a bunch of things at once and you figure out the best compromise, or if that stinks, you step back and choose different drivers. The real question in the OP was about a more specific issue: having a crossover point that reaches into the range where localization depends on phase. I gave my take on that.
The assumption that we should avoid a crossover in the 1.5-5kHz range is not one I agree with, but the question was "if you're not going to have a crossover there, is one below that a problem because of how localization works in the brain?". I say no.
The assumption that we should avoid a crossover in the 1.5-5kHz range is not one I agree with, but the question was "if you're not going to have a crossover there, is one below that a problem because of how localization works in the brain?". I say no.
There is one error in the linked page: We don't use phase shift for the sound localisation but delay. The difference might at first sound academic but it is important to know that one can perfectly calculate the phase shift for a given delay at a given frequency while it is not possibe to calculate a time delay from phase shift. The consequence is that localisation using phase-shift would very often result in an abiguous result while the use of delay gives accurate results in most cases.
Because crossovers introduce phase-shift and therefore group delay (i.e. a frequency dependant temporal dispersion of the input) the hearing is robbed of some accuracy in determining the interaural delay of an acoustic event. Because most recordings are intensity stereophony it may probably matter less. But the crossover frequency can definitley have an influemce on localisation. But that is just one parameter of many that have to be traded against the other ones in order to reach an optimum result with given drivers.
Regards
Charles
Because crossovers introduce phase-shift and therefore group delay (i.e. a frequency dependant temporal dispersion of the input) the hearing is robbed of some accuracy in determining the interaural delay of an acoustic event. Because most recordings are intensity stereophony it may probably matter less. But the crossover frequency can definitley have an influemce on localisation. But that is just one parameter of many that have to be traded against the other ones in order to reach an optimum result with given drivers.
Regards
Charles
All about the drivers. Distortion, dispersion, breakup modes, etc. The best answer is the best solution for the drivers in question. This is what Dumptruck is telling you.
There is considerable disagreement if we are sensitive to phase or just time/frequency variations. (outer ears doing directional equalization) Studies have proven both. The majority camp is in the time side. Truth is, a lot goes on in out brains we don't fully understand.
How important is this? Well I have built simple 2 ways with localization wider than the speakers and some semblance of height. Height being most likely careful equalization of certain tracks as that is what seems to provide height clues. I attribute this to being able to do a pretty good job of flat response in-room.
There is considerable disagreement if we are sensitive to phase or just time/frequency variations. (outer ears doing directional equalization) Studies have proven both. The majority camp is in the time side. Truth is, a lot goes on in out brains we don't fully understand.
How important is this? Well I have built simple 2 ways with localization wider than the speakers and some semblance of height. Height being most likely careful equalization of certain tracks as that is what seems to provide height clues. I attribute this to being able to do a pretty good job of flat response in-room.
Tvrgeek, it seems to me that Dumptruck is saying the opposite?
My first post above is a disagreement to why after reading the link the OP came to a conclusion that we should not cross below 1kHz (and better above 5kHz).
Dumptruck also said NO to the question.
The situation is now either me or Dumptruck didn't explain why NO. For me I think I know something but I don't want to discuss it because it is just some kind of "hypothesis" (and I'm here more into seeking new information than getting into useless debate).
So for the experts, if you want to explain the why not, go ahead
I think that large drivers (15") tend to have a ragged response at the top of their abilities. 5khz is pretty much beyond their abilities unless you have some magical woofer your not sharing with us? Now if your using say a 8" or 10" woofer then you would have a much better chance of pulling off a 5KHZ crossover because the smaller a driver gets the better it plays higher but the less it plays lower.
Last edited:
No, I'm in agreement with Tvrgeek. I was just saying that rehashing basic speaker design concepts is boring and unnecessary, The OP asked a specific question about localization being disturbed by phase shift in the crossover, when the crossover is low enough. Far as I can see, my post #8 and post #13 above are the only ones that actually tried to address the question.
OK
Let us see...
Isn't this also one of the basics you don't want to bother the OP with?
Whatever that may be?
I think without understanding that phase shift is different with group delay (but both are important consideration) good crossover design will be confusing.
Though the OP didn't expect discussing about the trade-off (which I assumed because he might think he knew enough), it might be necessary to iterate "warning". I did the same thing when I said that 1K is only a guideline but the exact optimum number will be found in the implementation. Gainphile for example, shown an example how he came up with 1K5.
I think 1K is only a guideline for those who do not understand why it has to be 1K, or don't understand how to make a good crossover at critical band (1K-5K). In practice, even 2K or slightly more is okay if you understand what to watch out for. And this is also from localisation point of view.
In my view, the localization issues are secondary to the distortion issues. Speakers are all so bad, it is a trade-off of major problems for the best overall compromise for each individual's preference.
I am talking practical drivers I can actually buy. If I could spec a "perfect" driver, then yes, I might cross outside the 1K to 5K region to get a continuity in group delay as some general practice. I can't. Most of my solutions using 1" tweeters wind up being around 2.6K. Seems like it is low enough so my mid-base is still reasonably well behaved, and high enough to not cause too much distortion from the tweeter. When my crossovers are done well, I get very good localization. The room treatments are more important at this point.
A major mistake taken by many very smart folks is to focus on a parameter, however real it may be, at the exclusion of far more important ones. Polk, Bose, and Klipsh all come to mind where they went too far on their favorite problem resulting in overall very bad results. The single driver camp premise is the lack of crossover giving better localization at the loss of almost everything else important. Vanderstein takes a more holistic approach, and the results show.
Now to get into localization a bit more, it is also closely related to equalization. Shifting the balance of multi-tone signals has the effect of moving them in space. This is a function of our ears. Shifting the eq right to left can also move the apparent image. Changing the reflection patterns in the room may move the apparent image.
I am talking practical drivers I can actually buy. If I could spec a "perfect" driver, then yes, I might cross outside the 1K to 5K region to get a continuity in group delay as some general practice. I can't. Most of my solutions using 1" tweeters wind up being around 2.6K. Seems like it is low enough so my mid-base is still reasonably well behaved, and high enough to not cause too much distortion from the tweeter. When my crossovers are done well, I get very good localization. The room treatments are more important at this point.
A major mistake taken by many very smart folks is to focus on a parameter, however real it may be, at the exclusion of far more important ones. Polk, Bose, and Klipsh all come to mind where they went too far on their favorite problem resulting in overall very bad results. The single driver camp premise is the lack of crossover giving better localization at the loss of almost everything else important. Vanderstein takes a more holistic approach, and the results show.
Now to get into localization a bit more, it is also closely related to equalization. Shifting the balance of multi-tone signals has the effect of moving them in space. This is a function of our ears. Shifting the eq right to left can also move the apparent image. Changing the reflection patterns in the room may move the apparent image.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.